Combining Augmented Analytics Time Series Forecasting Models with Planning

Objectives

After completing this lesson, you will be able to:

  • Combine an augmented analytics time series forecasting models with planning in SAP Analytics Cloud

Enhance a Business Plan with Predictive Model Outputs

Predictive Analytics Naturally Complements Planning

Predictive forecasts are based on actual historical data. These predictive forecasts are important extra sources of information for users such as financial controllers, supply chain planners, and sales representatives, to make decisions using their business acumen and knowledge.

The automated machine learning capabilities in SAP Analytics Cloud can assist users to enhance business decision-making.

Machine Learning Helps Planners Take Confident Decisions with SAP Analytics Cloud Predictive Planning

  • Predictive forecasting automates the creation of baseline planning models.
  • Planners then supplement these forecasts based on business acumen.
  • Planners monitor plan attainment based on continuously updated predictions.
  • SAP Analytics Cloud predictive planning supports top-down and bottom-up planning processes–enabling the planner to automatically build forecasts at the correct level of detail.

Predictive Forecasting Combines the Power of Planning and Predictive

Planners can use time series forecasting at large scale to support key use cases related to expense forecasting, sales planning, and/or workforce planning, for example. They can then complement the planning and budgeting activities with a data-driven approach to planning.

Planners benefit from Smart Predict capabilities to generate predicted forecasts, considering one or several business dimensions. Smart Predict allows them to analyze forecast accuracy by dimension value and understand signal breakdown in detail.

Planners select a planning model as the input of Smart Predict, generate segmented predictive models, and deliver predictive forecasts in the planning grid in a few clicks.

Your input is minimal: actuals are used by default. Only relevant columns are displayed and selection of multiple dimensions is supported for bottom-up forecasting.

Integrated Predictive Functions

  1. Augment time series charts and line charts with predictive forecasts:
  2. Augment the planning grid with forecasts:
  3. Create predictive models with greater levels of control using Smart Predict (including bottom-up time series models) and serve both SAP Analytics Cloud BI and planning needs:

Integrate a Time Series Model with Planning

Run a Predictive Time Series Forecast from a Planning Model

  1. Prepare the data by importing Actual data from a source system or from a flat file into an SAP Analytics Cloud planning model and creating a Forecast (or private) version.
  2. Create a table in a story and select a single cell that can be edited to apply a prediction to the specific selected measure.
  3. From the Tools options in the menu, select Predictive Forecast. If it is not visible in the menu, select ...More.
  4. Under Predictive Forecast Settings and Advanced Options, customize your forecast as required and select Preview.
    • Under Algorithm, select either Automatic Forecast, Linear Regression, or Triple Exponential Smoothing for the type of forecast that you want to create.
    • Under Output Values, select Positive Only so that no negative values are generated in the results.
    • Select which Version, to use for the forecast to input the data. By default, the Actual version is used.
    • Use Data From, specifies the start of the historical data time period the algorithm runs on.
  5. If you are happy with the preview, select OK to populate the table with the forecasted values.

Combine Augmented Analytics Time Series Forecasting Models with Planning

Scenario

You have been asked to create a time series forecasting model output with an SAP Analytics Cloud planning model.

The data you have been provided is as follows:

  • Monthly sales for a five-year period. This is the date variable.
  • Sales for six car brands in six countries. This is the target variable.
  • Not all brands are sold in all countries. This is used for entity definition.
  • Each country is associated with one of two regions (EMEA north and south). Region becomes a property of the Country.

The data is used as the input of a time series predictive scenario with the goal of forecasting the evolution of car sales segmented by country and brand for the next 12 months.

What skills do you develop in this practice exercise?

In this practice exercise, you perform the following tasks in SAP Analytics Cloud:

  1. Create planning model from CSV file
  2. Update planning model settings
  3. Create a story and private planning version
  4. Train forecasting models and inspect their quality
  5. Write forecasts back into planning model and use them in story

Task 1: Create Data Model

Task 2: Create a Story and Private Planning Version

Task 3: Train Forecasting Model in Smart Predict and Inspect Model Reports

Log in to track your progress & complete quizzes